[ruby-dev:50257] [Ruby trunk Bug#13917] Comparable#clamp is slower than using Array#min, max.

Issue #13917 has been updated by kei-s (Kei Shiratsuchi).


Thank you, Hanmac. I understand why `Array#min/max` is so fast.
I guess `clamp` would be implemented in `numeric.c` and so on to be as fast as `Array#min/max`.

----------------------------------------
Bug #13917: Comparable#clamp is slower than using Array#min,max.
https://bugs.ruby-lang.org/issues/13917#change-66816

* Author: kei-s (Kei Shiratsuchi)
* Status: Open
* Priority: Normal
* Assignee: 
* Target version: 
* ruby -v: 
* Backport: 2.2: UNKNOWN, 2.3: UNKNOWN, 2.4: UNKNOWN
----------------------------------------
Comparable#clamp is slower than using Array#min,max.
(I noticed it by @onk's tweet. https://twitter.com/onk/status/907856892604461056)

### Performance

```
              user     system      total        real
minmax:   0.740000   0.000000   0.740000 (  0.732744)
clamp:    2.060000   0.010000   2.070000 (  2.072794)
```

### Test Code

```
require 'benchmark'

Benchmark.bmbm do |x|
  v = Random.rand(-10..110)

  x.report "minmax:" do
    10000000.times { [99, [0, v].max].min }
  end

  x.report "clamp: " do
    10000000.times { v.clamp(0, 99) }
  end
end
```

# Patch

I made patch for it. But I'm not sure this is good way.
https://gist.github.com/kei-s/b303aca105df5c26be9c98f833db80f7#file-compar-diff

## After

```
              user     system      total        real
minmax:   0.820000   0.000000   0.820000 (  0.822517)
clamp:    1.090000   0.000000   1.090000 (  1.087491)
```

Other benchmark for this patch is here.
https://gist.github.com/kei-s/0c34cbe4e21a499601e8247077629082

## Questions

1. Should `clamp` version be faster than `Array#min/max` version?
`Array#min/max` version would have overhead of array creation.

2. Is `OPTIMIZED_CMP` in `cmpint` best way?
Some method doesn't pass `cmpint` (e.g. `Integer#>`). But `OPTMIZED_CMP` checks Integer. 




-- 
https://bugs.ruby-lang.org/